Overview

libmolgrid is a library to generate tensors from molecular data, with properties that make its output particularly suited to machine learning.

libmolgrid abstracts basic input generation functionality used in our related project, gnina, with applications reported in several papers including Protein-Ligand Scoring with Convolutional Neural Networks.

It’s implemented in C++/CUDA with Python bindings, with first-class integration for PyTorch, Caffe, and Keras.

Installation

Recommended:

  pip install molgrid

Install via conda is also supported:

  conda install -c jsunseri molgrid

Alternatively, you can build from source:

  pip3 install numpy pytest pyquaternion

Install cmake 3.12 or higher.

Install CUDA.

Install OpenBabel 3.0 (Not yet released, build from master).

  git clone https://github.com/gnina/libmolgrid.git
  cd libmolgrid
  mkdir build
  cd build
  cmake ..
  make -j8
  sudo make install